The relative abundances of short-range correlated (SRC) nucleon pairs in atomic nuclei are typically extracted from measurements of the per-nucleon inclusive electron scattering cross-section ratio of nucleus $A$ to the deuteron, $a_2=(\sigma_A/A)/(\sigma_d/2)$, at high-$x_B$ and large-$Q^2$. Despite extensive measurements, theoretical calculations of $a_2$ are sparse. Here we study the sensitivity of $a_2$ to the abundance of SRC nucleon pairs using the generalized contact formalism (GCF), which successfully describes nucleon knockout from SRC pairs up to 1000 MeV/c of initial momentum. The GCF reproduces the measured plateau in the cross-section ratio and the values of $a_2$. However, using a non-relativistic instant-form formulation, the data are only reproduced using model parameters that are inconsistent with ab-initio many body calculations. The calculations also show large sensitivity to the model parameters. Using a light-cone GCF formulation significantly decreases this sensitivity and improves the agreement with ab-initio calculations. We conclude that empirical SRC pair abundances extracted directly from $a_2$ measurements cannot be regarded as having accuracy that is better than $\sim 20\%$ and that precision studies of the nuclear mass and asymmetry dependence of SRC pair abundances must rely on cross-section calculations that properly account for relativistic effects.